TECH FOCUS - GREEN & CLEAN

Cleaning silicon wafers with an argon/nitrogen cryogenic aerosol process

James F. Weygand, Nat Narayanswami, and Daniel J. Syverson, FSI International

As IC processing technology becomes increasingly complex, circuits are becoming more sensitive to particles, residues, and damage from wet cleaning processes. There are inherent tradeoffs involved in wet chemical cleaning. Sulfuric-peroxide baths dissolve organic particles effectively, yet they leave sulfur residues and will not react with inorganic particles. A common megasonic cleaning solution, SC-1, oxidizes and etches device layers in order to lift off contaminants. Thin gate oxides are becoming increasingly susceptible to such chemical attack. At the same time, wafer size is increasing, making it more difficult to effectively rinse and dry wafers.

In an effort to deal with some of the limitations of wet cleaning, a recently commercialized CryoKinetic system (Aries, FSI International, Chaska, MN) offers an alternative cleaning method. Aerosol cleaning differs from other cleaning processes because it:

• Avoids the use of surface-modifying reactive chemicals.

• Does not require undesirable, solvent-based cleaning solutions.

• Requires no rinsing and drying, thus eliminating potential water spots.

• Uses only inert nitrogen and argon gases to clean.

These attributes could allow process engineers to use aerosol cleaning throughout the process line. Since its introduction, the aerosol process has addressed contamination control in etch, deposition, and other applications (see Table I).

Application Point	Technology Insertion	Purpose	Wet Clean Replacement
Dielectric etch	DRAM, logic	Scrap reduction	Yes
Sputter deposition	DRAM	Yield improvement	No
Gate metal silicides	DRAM	Reliability improvement	Yes
Metal etch	Advanced logic	Yield improvement	Yes
Via etch	DRAM, logic	Yield improvement	Yes
LPCVD deposition	DRAM	Yield improvement	No
Electrical probe	Logic	Yield improvement	Yes

Table I: Current applications of aerosol cleaning in manufacturing lines.

The argon/nitrogen (Ar/N^₂) aerosol successfully removes a wide variety of contaminants, including particles embedded in deposited films, reactor flakes, and etching residues. Because of the aerosol's chemically inert nature, it can be applied almost randomly to defect-prone areas. Much research has been conducted on understanding the effects of aerosol cleaning, with development work under way that focuses on other defect elimination and yield enhancement areas.

Development of Aerosol Cleaning

Researchers at IBM's T. J. Watson Research Center (Yorktown Heights, NY) looked at cleaning methods for silicon wafers that would be chemical-free and compatible with integrated processing in cluster tools, with the Ar/N^₂ aerosol cleaning technology resulting from a collaboration with Air Products and Chemicals.^1,2 A prototype tool was eventually built. To produce manufacturing-ready aerosol cleaning systems, IBM licensed the technology to FSI International, which refined the original design and offers the aerosol cleaning system to all chipmakers.³ This article describes the process and tool setup of the Aries system as well as results from various cleaning experiments.

Process Description

Figure 1 shows a simple block diagram of the tool. The aerosol clean process starts with argon and nitrogen gas. The volume of each gas is controlled individually by mass-flow controllers, and the gases are then mixed before entering the cryogenic chiller. The cryogenic chiller is a liquid nitrogen heat exchanger that cools the gas mixture to cryogenic temperatures. This cooled mixture flows to an injection nozzle, then is injected into the cleaning chamber through an array of nozzle holes. Expansion of the gas into the vacuum chamber, which is controlled at a partial vacuum of 100—300 torr, causes further cooling. This leads to the formation of frozen crystals as the gas temperature drops below 84 K, the triple point for argon (the point where the gas, liquid, and solid phases can coexist). Nitrogen in the gas mixture permits higher expansion ratios, increasing the cooling of the mixture without increasing argon usage. It also acts as a diluent for the argon. This provides a means for controlling argon particle size and kinetic energy. ¹

Figure 1: Simplified block diagram of CryoKinetic cleaning system.

The wafer passes under the nozzle and is sprayed with an aerosol consisting of the Ar/N^₂ mixture and argon ice crystals. The injector nozzle angle of incidence can be adjusted to optimize cleaning of high-aspect features. The intensity of the aerosol, energy, and size of the ice crystals are controlled by total gas flow, the ratio of argon to nitrogen, gas temperature, and chamber pressure.

During the cleaning process, the chamber pressure is maintained at 100—300 torr to prevent particles from settling in the chamber. The high-flow gas sweeps ice crystals out of the process chamber and into a heated pumping line where they are sublimated. Following cleaning, the chamber is pumped to base pressure, and any remaining gas is pumped out of the chamber. In the event a cleaning module is integrated into some other platform, the pressure would be equalized to the pressure of the host platform.

Cleaning Mechanism

The aerosol cleaning mechanism consists of argon ice crystals, cryogenic temperatures, and a high flow of argon and nitrogen gas. All three elements have a role in the cleaning process, although the importance of each is relative to what is being cleaned. The most important element, though, is the argon ice. These ice crystals can reach the wafer through the boundary layer formed by the gas blowing across the surface. The crystals have enough energy to overcome the adhesion forces holding submicron-sized particles to the surface. They are also capable of breaking up and removing residues or contaminants left on the surface by wafer processing.

The second element is the aerosol's cryogenic temperature. The extreme cold of the aerosol is believed to help separate dislodged contaminant particles from the surface through thermophoresis, which occurs when a particle, suspended in a gas with a temperature gradient, experiences motion toward the cooler region of the gas. The motion results in a net force on the particle caused by unequal momentum transfer from hot and cold gas molecules colliding with the particle. Since the gas adjacent to cold surfaces is cooler than in other regions, cold surfaces tend to attract particles, hot surfaces to repel them.

During the wafer cleaning, the surface is initially hot relative to the aerosol stream. This temperature gradient decreases with processing time as the wafer gets colder. However, the wafer temperature does not fall below that of the flow stream, so a positive thermophoretic force is maintained on particles throughout the cleaning process. The positive force enhances particle suspension, helping to remove them from the wafer surface (see Figure 2). These thermophoretic forces also inhibit redeposition of particles on the surface. The chuck is heated to maintain a high temperature gradient between the wafer and the aerosol. The cryogenic temperatures also help break up sidewall etch residues and other contaminant films. The impact of temperature will be greatest where there is a large difference in thermal conductivity between the wafer surface and the unwanted residue. The largest effect would likely be in removing polymeric etch residues from conductive films, such as polysilicon or metal lines.

Figure 2: The effect of temperature on particle suspension.

The third element is the flow of gas across the wafer surface. The primary role of the gas is to clear away the contaminants that have been removed from the surface. The gas flow carries these particles and residues out of the chamber, where they can be removed by the system vacuum pump. The gas will also provide the force to remove supermicron particles from the wafer surface. The high gas flow prevents redeposition on the wafer surface.

CryoKinetic Cleaning Results

Aerosol cleaning has successfully removed several different types of particles and residues, including particles embedded in PVD films and postdeposition particulates. It is being used, either alone or in conjunction with other cleans, at multiple process steps to reduce defects and improve yield.

Table II illustrates the effectiveness of the aerosol process in removing surface particulates. Silicon nitride contaminants were deposited on wafers and subjected to the CryoKinetic clean. The table plots removal efficiency versus particle diameter, and the results show that the aerosol process is effective over a wide range of particle sizes.

Particle Size (µm)	0.15­0.2	0.2­0.3	0.3­1.0	1.0­5.0	>5.0
Removal efficiency	96.5%	93.9%	98.5%	99.5%	97.7%
Standard deviation	±0.4%	±1.0%	±0.1%	±0.1%	±0.6%

Table II: Cleaning efficiency versus particle size.

Another example of particle removal has been demonstrated on wafers contaminated by a metal-silicide PVD process. Reactor particles deposit during the metal-silicide sputtering process, with many of them embedding in the film. Figure 3 shows a comparison of the aerosol clean to two different wet cleans, a sulfuric peroxide clean and an ammonia peroxide clean. Each of the wet cleans removed about 2% of the particles, probably those residing on the surface of the film. The aerosol clean removed more than 50% of the particles, the majority of which were partially embedded in the film. Implementing the aerosol clean at this step led to significant yield improvement.

Figure 3: Comparison of partially embedded metal-silicide particle removal capabilities.

Aerosol cleaning has also been used to remove reactive ion etch (RIE) residues, particularly after metal and polysilicon etch. Shrinking dimensions have created residues that are more difficult to remove with conventional methods. As a result, device manufacturers are very interested in alternative cleaning technologies that minimize the damage to metal films and underlying oxides. This is especially true for gate stack polysilicon etch, where the underlying oxide is the gate oxide itself.

Figure 4: SEM image of polysilicon etch residue.

Figure 5: SEM image of polysilicon lines after aerosol cleaning.

Figures 4 and 5 are scanning electron microscope (SEM) photographs of aerosol cleaning after gate polysilicon etch and ash. Before cleaning (Figure 4), there were prominent etch residues or fences along the polysilicon sidewalls. These residues were completely removed by the aerosol clean. Figures 6 and 7 show the effect of the aerosol clean on etch residues along aluminum lines. In Figure 6 the postetch residues are evident; Figure 7 shows the complete removal of these residues.

Figure 6: SEM image of aluminum etch residue.

Figure 7: SEM image of aluminum lines after aerosol cleaning.

A significant reliability issue faced by device manufacturers is post-RIE metal corrosion. Figure 8 depicts corrosion on etched aluminum lines on a wafer that was etched and ashed but not cleaned. Figure 9 shows a wafer etched and ashed at the same time, but aerosol cleaned 24 hours later. Both SEMs were taken one week after etching. The absence of corrosion on the aerosol-cleaned wafer indicates that the aerosol clean has removed the corrosion source.

Figure 8: SEM image of metal corrosion on aluminum lines on uncleaned wafer.

Figure 9: SEM image of aerosol-cleaned metal lines showing no corrosion.

A more subtle type of contamination has been found between metal conductors. Metal electrical test lines are arranged close to one another to measure intermetal shorting. Placing a voltage between the separate conductors should not produce any current flow. Figure 10 shows the results of such measurements taken on serpentine metal lines spaced 0.6 µm apart. Wafers 1 through 8 received the aerosol clean, and 9 through 16 received a standard wet clean. A very dramatic reduction in leakage current was seen on wafers cleaned with the aerosol cleaning process. This difference in leakage current was probably caused by surface films or residues not visible by SEM inspection.

Figure 10: Graph of line-to-line current leakage for wafers processed using aerosol and wet cleans.

Aerosol cleaning removes some contaminants better than others. The best results generally occur on particles with high profiles and low surface contact areas. Films that are more than several thousand angstroms thick can be removed with aerosol cleaning, although the processing time required to do so may make it impractical. The cryogenic temperatures of the aerosol can sometimes help break up films or reduce adhesive forces, making a difficult cleaning task less challenging. The aerosol clean is inherently versatile. For example, aerosol cleaning can be performed before or after ashing photoresist, or at some other intermediate step that may simplify contamination removal.

Possible Side Effects

Researchers have also investigated the possible side effects of aerosol cleaning. Particular attention has been given to patterned aluminum structures because they are thought to be the most sensitive to process-induced damage. Similar to the electrical test mentioned above, 0.6-µm serpentine lines were evaluated for changes in the individual line resistivity. Ideally, no change in resistance should occur if the line is not damaged. Figure 11 shows electrical test results for 16 wafers, half cleaned with the aerosol clean, the other half with a wet clean. Current levels were the same after the two processes, indicating that the aluminum structures were not damaged by either cleaning process.

Figure 11: Graph showing aerosol cleaning did not cause any degradation of line resistance.

Potential surface charging was also studied. Dielectric test structures with 100-Å gate oxide were tested for gate oxide integrity following an aerosol clean. Each wafer had 74 die with three large, antenna-ratio capacitors. The ratios of polysilicon to gate areas were 0.2 million, 1.0 million, and 10 million to 1, respectively. The yields listed in Table III are expressed as the percentage of die per wafer that exhibit leakage current <1 µA during a 0—30-V ramp at 0.2 V/sec. No failures were detected, indicating no charge-induced damage to sensitive gate oxides.

Sample	Polysilicon: Gate Oxide Antenna Area
	0.2 million:1	1.0 million:1	10.0 million:1
Control	100%	100%	100%
Aerosol-cleaned wafer	100%	100%	100%

Table III: Percent of chips per wafer with <1-µA leakage current in electrical tests of gate oxide integrity.

Surface roughening effects on bare silicon wafers were assessed using atomic force microscopy (AFM). Wafers were subjected to multiple passes under the aerosol stream well in excess of normal operating conditions. The results shown in Figure 12 indicate that there was no surface roughening when the aerosol clean process was used.

Figure 12: Surface roughness of bare silicon wafers treated to multiple aerosol passes.

Sample	Mean (Å)	Std. Dev. (%)	Min. (Å)	Max. (Å)
Control	595.37	0.557	586.57	600.90
Control repeat	595.47	0.573	586.46	601.08
Before clean	586.98	0.595	578.20	593.13
After clean	587.17	0.591	578.39	592.70

Table IV: Change in oxide film thickness and roughness after aerosol cleaning.

Another measurement of potential surface damage is blanket oxide thickness after aerosol cleaning. Table IV contains the results of oxide-thickness measurements on two wafers. One wafer is a control and was measured twice to look at measurement repeatability, with the second wafer measured before and after aerosol cleaning. The results show that the aerosol clean produced no discernible change in oxide thickness or roughness.

Cross-Contamination

One must understand the possible cross-contamination effects of any cleaning procedure. Processing dirty wafers raises the concern that particulate debris can carry over to subsequent wafers in a sequential process. To understand this effect, particle changes were evaluated on clean wafers processed immediately after highly contaminated ones. Before cleaning was begun, the background contamination levels on the system were measured by processing clean monitor wafers. Figure 13 shows the range and average change on 200-mm wafers, with the data representing all particles >0.20 µm. As the results show, the system does not add any background contamination.

Figure 13: Background particle data collected prior to cross-contamination experiments.

The contaminated-wafer study was then conducted. Contaminated wafers used in this test were blank silicon wafers coated with photoresist and then partially ashed, so that about 90—95% of the original resist film was removed. There were three groups in the experiment. First, four baseline particle monitors were run, repeating the previous test. Second, one contaminated wafer was cleaned, followed by three particle monitors placed throughout the cassette in adjacent positions and separated by dummy wafers. Third, eight contaminated wafers were cleaned, followed by particle monitors as in Test 2. The results of this test, shown in Table V, indicate that there is no carryover of particles in the system.

Test No.	Avg. Preclean	Avg. Postclean	Avg. Delta
1	33	30	-3
2	30	35	5
3	100	33	-67

Table V: Effect of process recipe and running contaminated wafers on particle adders from cryogenic cleaning.

Conclusion

The results of the various investigations demonstrate that aerosol cleaning can remove a wide range of contaminants found in semiconductor processing lines. The aerosol clean process uses no chemicals and appears to be safe for use in all areas of wafer processing. The vacuum nature of the process shows great potential for meeting future cleaning integration needs. Experiments designed to uncover side effects of the cleaning process indicate no degradation of sensitive structures, suggesting that the process can be readily inserted in a variety of steps in the semiconductor manufacturing process.

Acknowledgments

The authors want to thank the researchers at IBM's Advanced Semiconductor Technology Center in Hopewell Junction, NY, for providing much of the data and other information reported on in this article.

References

1. McDermott WT, Ockovic RC, Wu JJ, and Miller RJ, "Surface Cleaning by a Cryogenic Argon Aerosol," in Proceedings of 37th Annual Technical Meeting of the Institute of Environmental Sciences, Mount Prospect, IL, IES, pp 882—885, 1991.

2. McDermott WT, Ockovic RC, Wu JJ, and Miller RJ, "Removing Submicron Surface Particles Using a Cryogenic Argon-Aerosol Technique," Microcontamination, 9(10):33—36, 94, 95, 1991.

3. Wu JJ, Syverson D, Wagener T, and Weygand J, "Wafer Cleaning with Cryogenic Argon Aerosols," Semiconductor International, 19(9):113—118, 1996.

James F. Weygand joined FSI International, Chaska, MN, as an applications engineer in the company's surface conditioning division in 1995. He began his career at IBM in East Fishkill, NY, in 1967, then transferred to IBM's facility in Rochester, MN, in 1971 to help set up an IC engineering pilot line. Weygand held process engineering positions in photolithography, mask build and thin films, and process characterization in the Rochester IC line from 1971 to 1989, when he became engineering manager for the line. He has a BS in physics (1967) from Polytechnic Institute of Brooklyn, and MS degrees in electrical engineering (1981) and management of technology (1994) from the University of Minnesota.

Nat Narayanswami, PhD, is a senior engineer in the analysis group at FSI International. He was a research associate at Tohoku University in Sendai, Japan, before joining the company in 1994. Narayanswami holds a PhD in mechanical and aerospace engineering (1992) from Rutgers University. He is a member of the American Society of Mechanical Engineers, American Physical Society, and American Institute of Aeronautics and Astronautics and has authored more than 15 published papers. (Narayanswami can be reached at 612/448-1334; e-mail, nnarayanswami@fsi-intl.com)

Daniel J. Syverson is marketing technologist for single-wafer surface conditioning products at FSI International. He joined the company in 1985 as an applications engineer and was later promoted to applications and design engineering manager for vapor products. Syverson worked at Honeywell's solid state electronics center between 1979 and 1985, where he held various process engineering and development positions. He has an AAS in electronics technology from Brown Institute and is enrolled in the business management and marketing program at Concordia College (St. Paul, MN). (Syverson can be reached at 612/448-8048.)

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit